# Axiomatic Systems for Geometry

Shared by:
Categories
-
Stats
views:
25
posted:
7/29/2010
language:
English
pages:
8
Document Sample

```							               Axiomatic Systems for Geometry∗

George Francis†

15 January 2002

1      Basic Concepts

An axiomatic system contains a set of primitives and axioms. The primitives
are object names, but the objects they name are left undeﬁned. This is why
the primitives are also called undeﬁned terms. The axioms are sentences that
make assertions about the primitives. Such assertions are not provided with
any justiﬁcation, they are neither true nor false. However, every subsequent
assertion about the primitives, called a theorem, must be a rigorously logical
consequence of the axioms and previously proved theorems. There are also
formal deﬁnitions in an axiomatic system, but these serve only to simplify things.
They establish new object names for complex combinations of primitives and
previously deﬁned terms. This kind of deﬁnition does not impart any ‘meaning’,
not yet, anyway.

If, however, a deﬁnite meaning is assigned to the primitives of the axiomatic
system, called an interpretation, then the theorems become meaningful asser-
tions which might be true or false. If for a given interpretation all the axioms
are true, then everything asserted by the theorems also becomes true. Such an
interpretation is called a model for the axiomatic system.

In common speech, ‘model’ is often used to mean an example of a class of things.
In geometry, a model of an axiomatic system is an interpretation of its primitives
for which its axioms are true. Since a contradiction can never be true, an axiom
system in which a contradiction can be logically deduced from the axioms has
no model. Such an axiom system is called inconsistent. On the other hand, if an
abstract axiom system does have a model, then it must be consistent because
each axiom is true, each theorem is a logical consequence of the axioms, and
∗ From Post-Euclidean Geometry: Class Notes and Workbook, UpClose Printing & Copies,
Champaign, IL 1995
† Prof. George K. Francis, Mathematics Department, University of Illinois, 1409 W. Green

St., Urbana, IL, 61801. (C)2202 Board of Trustees.

1
hence it is true, and a contradiction cannot be true.

Finally, an axiom system might have more than one model. If two models of the
same axiom system can be shown to be structurally equivalent, they are said to
be isomorphic. If all models of an axiom system are isomorphic then the axiom
system is said to be categorical. Thus for a categorical axiom system one may
speak of the model; the one and only interpretation in which its theorems are
all true.

All of these qualities: truth, logical necessity, consistency, uniqueness were tac-
itly believed to be the hallmark of classical Euclidean geometry. At the start of
the 19th century, a scant 200 years ago, philosophers and theologians, physicists
and mathematicians were all persuaded that Euclidean geometry was absolutely
the one and only way to think about space, and therefore it was the job of ge-
ometers to develop their science in such a way as to demonstrate this necessary
truth. By the end of the century, this belief had been thoroughly discredited
and abandoned by all mathematicians.1

The main theme of our course concerns the evolution of this idea, and its re-
placement by the much richer, far more illuminating, post-Euclidean geometry
of today. It is about a method of thought, called the axiomatic method. Al-
though at one time this method may have developed merely from a practical
need to verify the rules obtained from the careful observation of physical exper-
iments, this changed with the Greek philosophers. The axiomatic method has
formed basis of geometry, and later all of mathematics, for nearly twenty-ﬁve
hundred years. It survived a crisis with the birth of non-Euclidean geometry,
and remains today one of the most distinguished achievements of the human
mind.

As we noted earlier, the transition of geometry from inductive inference to
deductive reasoning resulted in the development of axiomatic systems. Next,
we look at four axiom systems for Euclidean geometry, and close by constructing
a model for one of them.

2     Euclid’s Postulates:

Earlier, we referred to the basic assumptions as ‘axioms’. Euclid divided these
assumptions into two categories — postulates and axioms. The assumptions
that were directly related to geometry, he called postulates. Those more related
to common sense and logic he called axioms. Although modern geometry no
longer makes this distinction, we shall continue this custom and refer to axioms
for geometry also as ‘postulates’.
1 Curiously, it persists even today among some irresponsible, but inﬂuential amateurs. See

“Ask Marilyn”, by Marilyn Vos Savant, Parade Magazine, November 21, 1993.

2
Here is a paraphrase2 of the way Euclid expressed himself.

Let the following be postulated:

Postulate 1: To draw a straight line from any point to any point.
Postulate 2: To produce a ﬁnite straight line continuously in a straight line.

Postulate 3: To describe a circle with any center and distance.
Postulate 4: That all right angles are equal to one another.
Postulate 5: That, if a straight line falling on two straight lines makes the
interior angles on the same side less than two right angles, the two straight
lines, if produced indeﬁnitely, meet on that side on which the angles are
less than two right angles.

Note that the wording suggests construction. Euclid assumed things that he
felt were too obvious to justify further. This caused his axiomatic system to
be logically incomplete. Consequently, other axiomatic systems were devised in
an attempt to ﬁll in the gaps. We shall consider three of these, due to David
Hilbert, (1899), George Birkhoﬀ, (1932) and the School Mathematics Study
Group (SMSG), a committee that began the reform of high school geometry in
the 1960s.

3           Hilbert’s Postulates:

In the late 19th century began the critical examinations into the foundations of
geometry. It was around this time that David Hilbert(1862 - 1943) introduced
his axiomatic system. The primitives in Hilbert’s system are the sets of points,
lines, and planes and relations, such as
incidence:            as in ‘a point A is on line ’
order:                as in ‘C lies between points A and B’
congruence:           as in ‘ line segments AB ∼ A B ’
=
An example of a formal deﬁnition would be that of a line segment AB as the set
of points C between A and B. He partitioned his axioms into ﬁve groups; ax-
ioms of connection,order, parallels, congruence and continuity.3 Hilbert’s axiom
system is important for the following two reasons. It is generally recognized as a
ﬂawless version of what Euclid had in mind to begin with. It is purely geomet-
rical, in that nothing is postulated concerning numbers and arithmetic. Indeed,
it is possible to model formal arithmetic inside Hilbert’s axiomatic system.
2 Thomas     L. Heath, “The Thirteen Books of Euclid’s Elements”, Cambridge, 1908.
3 cf.   Wallace and West, op.cit., Chapter 2 for a more detailed discussion of Hilbert’s axioms.

3
We wish to show how Euclidean geometry can be modelled inside arithmetic.4
For this purpose, we want the shortest possible list of primitives and postulates,
for then, we have less to check. Birkhoﬀ meets this requirement.

4     Birkhoﬀ ’s postulates

The primitives here are the set of points, a system of subsets of points called
lines, and two real-valued functions, ‘distance’ and ‘angle’. That is, for any pair
of points, the distance d(A, B) is a positive real number. For any ordered triple
of points A, Q, B, the real number m AQB is well deﬁned5 modulo 2π.

Euclid’s Postulate: A pair of points is contained by one and only one line.
Ruler Postulate: For each line there is a 1:1 correspondence between its points
and the real numbers, in such a way that if A corresponds to the real
number tA and B corresponds to tB then
d(A, B) = |tB − tA |
.
Protractor Postulate: For each point Q, there is a 1:1 correspondence of
its rays6 and the real numbers7 modulo 2π, in such a way that if ray r
corresponds to the circular number ωr and ray s to ωs then
m RQS = ωs − ωr (mod2π)
where R is a second point on r and S on s .
simSAS Postulate: If m P QR = m P Q R and d(P Q) : d(P Q ) = d(QR) :
d(Q R ) = k then the other four angles are pairwise equal, and the remain-
ing side pairs have the same ratio.

One says that such triangles are similar, P QR ∼               P Q R with scaling factor
k. Of course, for k = 1, P QR ∼ P Q R .
=
4 The historical signiﬁcance of these two exercises in building models of formal systems is

the irrefutable demonstration that geometry and arithmetic are equi-consistent. That means,
if you believe the one to be without contradiction, then you are obliged to accept the other
also, and vice-versa. Hilbert’s program for a proof that one, and hence both of them are
o
consistent came to naught with G¨del’s Theorem. According to this theorem, any formal sys-
tem suﬃciently rich to include arithmetic, for example Euclidean geometry based on Hilbert’s
axioms, contains true but unprovable theorems.
5 To distinguish the ﬁgure AQB, which we call an ‘angle’, the number m AQB is called

the angular measure of the angle. Moreover, two real numbers that diﬀer by a multiple of 2π
measure the same angle.
6 Note that once we can apply a ruler to a line, we can identify one of the two half-lines,

or rays, at a point Q as those points P on the line for which tP > tQ .
7 We might call these the circular numbers because they lie on the number circle, just as

one speaks of the real numbers lying on number line.

4
5       The SMSG Postulates

There are 22 of these,8 and they combine the ﬂavor of Hilbert and Birkhoﬀ. With
Birkhoﬀ, rulers and protractors are postulated, under the valid impression that
children already know how to deal with real numbers by the time they study
geometry. There are many postulates so that proofs of interesting theorems
can be constructed without the tedium of proving hundreds of lemmas ﬁrst. Of
course, unlike Birkhoﬀ’s foursome, the SMSG postulates are redundant, in that
some postulates can be logically derived from others. The pedagogical wisdom
and usefulness of the SMSG axiom system is a matter of some debate among
educators.

6       A Cartesian Model of Euclidean Geometry

We next give an example of an axiomatic system and a model for it. For this
purpose we choose a very familiar area of mathematics in which to interpret
the primitives and to test the truth of the axioms. We all know analytic plane
geometry from high school, also known as Cartesian geometry. Birkhoﬀ’s four
postulates for Euclidean geometry appear compact enough for us not to lose our
way.

We interpret the points A, B, C... as ordered pairs, (x, y), of real numbers. Lines
shall be solution sets to linear equations of the form ax + by + c = 0. A point
(p, q) is incident to the line ax + by + c = 0 if it satisﬁes the equation, i.e. if
ap + bq + c = 0 is true. Remember that the distance between two points and
the angle measure are also primitives and need an interpretation. We shall do
that later.

With just this much we can already attempt to verify the ﬁrst postulate which
asserts the existence and uniqueness of a line through two given points. You
could do this yourself by deriving the formula for the line through two points
(x0 , y0 ), (x1 , y1 ) in any of the many ways you learned to do this in high school.
Here we do this by solving this system of two linear equations for the as yet
unknown parameters a, b, c :

ax0 + by0 + c        = 0
ax1 + by1 + c        = 0
a(x1 − x0 ) + b(y1 − y0 ) = 0

The third equation eliminates c for the moment; we can recover it as soon as
we know a, b, for example thus:

c = −ax0 − by0 .
8 Cf.   Appendix of Wallace and West, op.cit.

5
One plausible choice for a,b would be

a = −(y1 − y0 ), b = (x1 − x0 )

because it ﬁts the third equation and yields

x0     y0
c = x0 y1 − x1 y0 =
x1     y1

While this shows that both points lie on some line, it does not demonstrate the
uniqueness of this line. Indeed, our interpretation is incomplete. If we really
want the ﬁrst postulate to hold we must agree that the same line may have more
than one equation, provided the same set of points is the solution set for each.
We therefore amend our interpretation of a line by stressing that

a0 x + b0 y + c0 = 0
a1 x + b1 y + c1 = 0

deﬁne the same line provided the parameters are proportional:

a0 : a1 = b0 : b1 = c0 : c1 .

The distance function d(A, B) Birkhoﬀ had in mind is, of course, the Euclidean
distance as derived from the Pythagorean theorem:

For A = (x0 , y0 ), B = (x1 , y1 ),

d(A, B) =        (x1 − x0 )2 + (y1 − y0 )2

We are now ready to verify Birkhoﬀ’s ruler postulate in a particularly useful
fashion. First we give our two arbitrary points more mnemonic names: Q =
(x0 , y0 ), I = (x1 , y1 ). Now there is a canonical way of labelling all other points
on the line QI with real numbers t in several useful ways as follows:

xt            x0                      x1
=           (1 − t) + t
yt            y0                      y1

In vector notation this might be written as

Pt = Q(1 − t) + tI = Q + t(I − Q).

Notice that P0 = Q and P1 = I, and that the points on the segment QI are
given by the set
{Pt |0 < t < 1}.

6
Problem 1: Recall that these are called parametric equations for the line; the non-parametric
equation is obtained by elimination of the t from the system of two linear equations. Verify
this.

There is still something to prove here, namely that the Euclidean distance is in
fact measured by our ruler. Once again we were too hasty in ruling lines. For
the Euclidean distance

d(Q, I) =      (x1 − x0 )2 + (y1 − y0 )2

which need not equal the parametric distance, which is 1. We may, however,
rescale our ruler by a unit u = d(Q, I), to yield another parameter, s = tu, for
which the points Ps on the line are given by Ps = Q + sU , where U is the unit
vector, U = (I − Q)/d(I, Q), in the direction if I from Q . This way, I is the
correct distance, d(I, Q), away from Q on this ruler for the line.

For the remaining pair of Birkhoﬀ’s postulates we need a protractor, i.e. a
device for measuring angles. The simplest way of doing this in our model is to
recall the deﬁnition of the dot product of two vectors and interpret:
A−Q B−Q
m AQB = arccos(               ·        ).
d(A, Q) d(B, Q)

Problem 2: Show that with this interpretation, Birkhoﬀ’s protractor postulate is true.

Birkhoﬀ’s axiom system achieved its remarkable economy by postulating what
turns out to be the quintessential property of Euclidean geometry. What distin-
guishes it from non-Euclidean geometry are the properties of geometric similar-
ity. Two shapes are similar if they diﬀer only in scale. Birkhoﬀ postulates that
two triangles with a similar corner, are wholly similar. By a corner we mean, of
course, a vertex and its adjacent edges. If the proportionality factor is 1, then
this postulate says that two triangles are congruent as soon as they have one
congruent corner.

We shall verify the simSAS postulate, which makes an assertion about two
triangles, by carefully measuring one triangle. Just as today we exchange goods
by means of their price, instead of bartering items for each other, so modern
geometry compares shapes by comparing their measurements.

Given a triangle     ABC, vital statistics consists of six numbers, the three angles
and sides,
α=m A
β=m B
γ=m C
a = d(B, C)
b = d(C, A)
c = d(A, B).

7
The law of cosines, which generalizes the Pythagorean theorem to arbitrary
triangles by resolving the square of a side in terms of the opposite corner:

c2 = a2 + b2 − 2ab cos γ.

allows us to measure c in terms of the measures of two sides and the included
angle.
Problem 3: Use vector algebra and the deﬁnition of the dot product to verify the law of
cosines. Hint: Multiply out
C 2 = (B − A)2 = ((B − C) − (A − C))2 .

Thus, knowing a, b and γ, we calculate c. If a and b are stretched, or shrunk by
the same factor, so is c, provided γ remains the same.
Problem 4: Apply the law of cosines to the other two sides to calculate α and β as functions
of a, b and γ.

Problem 5: A generalization of Euclid’s proof of the Pythagorean theorem leads to another
proof of the law of cosines. Label an arbitrary acute triangle in the standard way. Construct
squares on two of its sides, say b and c. Extending the altitudes from C and B partitions the
squares into rectangles
b2 = bb1 + bb2
c2 = cc1 + cc2
Euclid’s argument (do it!) proves that cc1 = bb2

Now drop the third altitude from A. Of course (can you prove this?) it passes through the
same point where the ﬁrst two altitudes intersected,9 and partitions the third square into two
rectangles.

Finally, we can measure the rectangle,    summarize our inferences and come up with the law of
cosines.
aa2 =        bb2 = ab cos C
c2 =       cc1 + cc2
=       bb2 + aa1
=       (b2 − bb1 ) + (a2 − aa2 )
=       b2 + a2 − 2ab cos C

Problem 6: Generalize the above argument to work also for an obtuse triangle. Hint:
Sometimes you need to add instead of subtract and vice-versa.

9 This   point is called the orthocenter of the triangle.

8

```
Related docs